

	
	u "data/panel_mafiaconsmig", clear
	
	drop if iddecade==1|iddecade==4 // 1951 / 1981

	
**-----------------------------------------------------
			
	* Appendix G: Additional predictions of the theory

**-----------------------------------------------------

*Table G.1: Controllable resources (migrants under and not under criminal control) and 
*demand for resources mafias can offer (sectors not requiring unskilled informal labor)

	*Migr North
	preserve 
		drop zmigshare zem zssmig zss 
		rename (zmigsharenorth zemnorth zssmignorth zssnorth)(zmigshare zem zssmig zss )
		xtivreg2 newspc dec2-dec3 ( zempcp zmigshare zem = zssc zssmig zss) ipop, fe first savefirst cluster(municipalities)
		mat t=e(first)
		local fs=round(t[4,3], 0.01)
		local fstat=round(t[8,3], 0.01)
		local ar=round(e(archi2), 0.1)
		outreg2 using "tables/tableG1", tex replace label dec(3) keep(zempcp zmigshare zem) addtext(City Decade FE, Yes) adds(F-Stat, `fs', SW F-Stat, `fstat', A-R Wald test, `ar') nocons nor2 ctitle("Migr not", "under mafia", "control")
	restore
	
	
	*Migr origin mafia
	preserve 
		drop zmigshare zem zssmig zss 
		rename (zmigshare_mafia zem_mafia zssmig_mafia zss_mafia)(zmigshare zem zssmig zss )
		xtivreg2 newspc dec2-dec3 ( zempcp zmigshare zem = zssc zssmig zss) ipop, fe first savefirst cluster(municipalities)
		mat t=e(first)
		local fs=round(t[4,3], 0.01)
		local fstat=round(t[8,3], 0.01)
		local ar=round(e(archi2), 0.1)
		outreg2 using "tables/tableG1", tex label dec(3) keep(zempcp zmigshare zem ) addtext(City Decade FE, Yes) adds(F-Stat, `fs', SW F-Stat, `fstat', A-R Wald test, `ar') nocons nor2 ctitle("Migr from", "mafia-affected", "province")
	restore
	
	*Not labor intensive, requiring local workforce (retail commerce)
	preserve 
		drop zempcp zem zssc zss 
		rename (zempcp_commerce zem_commerce zssc_traditional zss_traditional)(zempcp zem zssc zss)
		label var zempcp "Share Employed"
		xtivreg2 newspc dec2-dec3 ( zempcp zmigshare zem = zssc zssmig zss), fe first savefirst cluster(municipalities)
		mat t=e(first)
		local fs=round(t[4,3], 0.01)
		local fstat=round(t[8,3], 0.01)
		local ar=round(e(archi2), 0.1)
		outreg2 using "tables/tableG1", tex label dec(3) keep(zempcp zem zmigshare) addtext(City Decade FE, Yes) adds(F-Stat, `fs', SW F-Stat, `fstat', A-R Wald test, `ar') nocons nor2 ctitle("Not labor","intensive", "sector")
	restore
	
	*Booming regulated sector (petrolchem and car manufacturing)
	preserve 
		drop zempcp zem zssc zss 
		rename (zempcp_mechanic zem_mechanic zssc_traditional zss_traditional)(zempcp zem zssc zss )
		xtivreg2 newspc dec2-dec3 (zempcp zmigshare zem  = zssc zssmig zss), fe first savefirst cluster(municipalities)
		mat t=e(first)
		local fs=round(t[4,3], 0.01)
		local fstat=round(t[8,3], 0.01)
		local ar=round(e(archi2), 0.1)
		outreg2 using "tables/tableG1", tex label dec(3) keep(zempcp zmigshare zem) addtext(City Decade FE, Yes) adds(F-Stat, `fs', SW F-Stat, `fstat', A-R Wald test, `ar') nocons nor2 ctitle("Highly","regulated","industries")
	restore
	
	*High skilled sectors (financial services, insurance, press and editory, entertainment, communications)
	preserve 
		drop zempcp zem zssc zss  
		rename (zempcp_highskill zem_highskill zssc_traditional zss_traditional)(zempcp zem zssc zss)
	label var zempcp "Share Employed"
	label var zem "Emp x Migr"
		xtivreg2 newspc dec2-dec3 ( zempcp zmigshare zem = zssc zssmig zss), fe first savefirst cluster(municipalities)
		mat t=e(first)
		local fs=round(t[4,3],0.01)
		local fstat=round(t[8,3], 0.01)
		local ar=round(e(archi2), 0.1)
		outreg2 using "tables/tableG1", tex label dec(3) keep(zempcp zmigshare zem) addtext(City Decade FE, Yes) adds(F-Stat, `fs', SW F-Stat, `fstat', A-R Wald test, `ar') nocons nor2 ctitle("Sectors w","high-skilled","workforce")
	restore
	
	
	
*Figure G.1: Fewer small firms in places with larger construction boom

	*Y =Share of employers per number of firms
	* the smaller this quantity, the more construction industry in a city is based on few employees per firm
	* ranges from 1 (1 emp per 1 firm) to 420 (420 emp per firm) -- in the regression we normalize it
	gen emp_per_firm=(emp_c/firm_c)
	binscatter change_emp_cshare emp_per_firm, ytitle(Growth in construction employment) xtitle(Employer per firm)
	graph export "figures/figureG1.pdf", as(pdf) replace
	
	
*Table G.2: Benefits from the agreement (Effect of mafia infiltration on average firm size)
	*Gen dummy for cities with and without mafias at any point between 1960-1980
	bys municipalities: egen anymafia=max(newspc)
	gen anymafianews=(anymafia>0 & anymafia!=.)
	drop anymafia
	
	*Normalize emp_per_firm to range between 0 and 1
	qui sum emp_per_firm
    gen emp_per_firm01 = (emp_per_firm - r(min)) / (r(max) - r(min))
	
	label var emp_per_firm01 "Worker per firm ratio"
	label var anymafianews "Mafia presence"
	label var empci51 "Emp Constr 1951"
	label var Mdsouth55 "Migr South 1955"
		
	* test agreement
	
	su emp_per_firm
	local meandv=round(r(mean), 0.001)
	reghdfe emp_per_firm (zempcp c.zempcp#anymafia) zssmig $controls, abs(municipalities iddecade) cluster(municipalities)
	outreg2 using "tables/tableG2", tex replace label dec(3) adds(Mean DV, `meandv') addtext(F-stat, -) nor2 nocons ctitle("OLS")
	
	ivreghdfe emp_per_firm (zempcp c.zempcp#anymafia = zssc c.zssc#anymafia) zssmig ipop, abs(municipalities iddecade) cluster(municipalities) first
	local fstat=e(first)[4,1]
	outreg2 using "tables/tableG2", tex label dec(3) adds(Mean DV, `meandv', F-stat, `fstat') nor2 nocons ctitle("2SLS")
	